Evaluating the effectiveness of a multi-component intervention on early childhood development in paediatric HIV care and treatment programmes: a randomised controlled trial

The overall aim of this trial was to determine the impact of a comprehensive community-based intervention set up to simultaneously enhance child stimulation, reduce economic insecurity and improve retention in care among HIV exposed and infected children aged 0–24 month.

The specific aims of the trial include:

This trial was a collaborative project run by researchers from University College London, Centre for Sexual Health and HIV /AIDS Research (CeSHHAR) Zimbabwe, Liverpool School of Tropical Medicine, London School of Hygiene and Tropical Medicine, World Education, Stellenbosch University and Oxford University. CeSHHAR Zimbabwe was responsible for implementing and conducting all the research-oriented activities. World Education was responsible for providing technical support and ensuring delivery of the intervention by their implementing partner, Mavambo Orphan Care, as well collection of routine monitoring and evaluation data.

The trial was set in 30 health facilities in two districts in rural Zimbabwe, Goromonzi (n = 20) and Mudzi (n = 10). Eligible health facilities were those providing and initiating maternal and paediatric ART with at least 30 eligible caregiver-child dyads on their HIV-exposed babies register. Health facilities were selected for the trial based on mapping of all facilities within the district and were at least 15 km apart to minimise the risk of contamination.

The trial population comprised of primary caregivers (biological and non-biological), and their HIV positive or exposed infants aged 0–24 months. Dyads must have been living in the catchment areas of designated trial health care facilities, planning to live full time in the community for the 18 months trial duration, willing to attend meetings in the community once every two weeks, and prepared to provide their residential address and other locator information for follow-ups. Only singleton births were eligible for inclusion. Dyads that included children with chronic illness, not including HIV exposed infants, or children with mental/physical disabilities as recorded by the health worker were excluded.

A 3-month pilot was conducted prior to the trial (August to October 2015). We recruited 50 caregiver-child dyads seeking health services in five clinics excluded in the trial and geographically separated from the trial clinics. The aim was to assess the acceptability and feasibility of the intervention, to determine feasibility and procedures for trial recruitment, examine preliminary data and to assess the relevance and appropriateness of the parenting programme content adapted to HIV affected families. The pilot resulted in minor changes to the finalised design of the trial and informed the final protocol. These changes included reducing ECS content per module, and increasing modules from 12 to 18, to be conducted fortnightly as opposed to proposed weekly or monthly. Cluster sizes were changed from an initial 16 caregiver-child dyads to either 12 or 24 caregiver-child dyads per cluster to cater for manageable Internal Savings and Lending Scheme (ISALS) composition. These changes informed the final study protocol.

Thirty health facilities and their surrounding catchment areas were randomised to the Child Health Intervention for Development Outcomes (CHIDO) intervention or standard of care (Fig. 2). Randomization of the 30 selected clusters was conducted with minimisation on the number of HIV exposed infants (0–24 months) in the Exposed Infant Registers, to ensure that the number of eligible caregiver-child pairs was similar by arm.

A list of eligible infants was extracted from the ‘Exposed Infant Registers’, comprising the name of the biological mother, village, name of the child, date of birth of the child, ART number of the mother and contact details. For the 30 trial clinics there were a total of 1509 eligible infants extracted from the health facility PMTCT registers. Case care workers (CCW) in catchment areas of villages with the highest number of clinics were asked to invite all eligible caregiver-child dyads who met the trial inclusion criteria (n = 671) to attend an orientation meeting to learn about the trial. If the biological mother listed in the register was no longer the primary caregiver, the non-biological caregiver was invited to participate as long as the child was listed as HIV exposed. Eligible caregivers who provided verbal consent to participate were booked for enrolment procedures – n = 574 and 97 (14.5%) declined enrolment. Enrolment in the CHIDO intervention and usual care arms occurred concurrently, with one clinic from each arm recruited at a time for implementation purposes. Enrolment took 6 months to complete. Follow up for each pair of clinics was 12 months after the CHIDO intervention arm has started receiving the intervention in their community.

All eligible caregiver-child dyads were given trial information and asked to provide written informed consent to participate. Informed consent was collected according to Good Clinical Practice guidelines.

Table 1 summarises the scheduled trial timeline, from participants’ enrolment, intervention roll out, follow up assessments to data analysis phase.

The primary outcome measures were:

The age-standardized Early Learning Composite (ELC) score was assessed for all infants at baseline and endline. Trained assessors (independent of the implementers) used the Mullen Scales of Early Learning focusing on the four cognitive scales (visual reception, fine motor, receptive language, and expressive language). The Mullen Scales of Early Learning Tool is an individually administered comprehensive measure of cognitive functioning for infants and preschool children from birth through 68 months [34]. Scores were adjusted based on the child’s age. The number of assessors was kept to a minimum of 3, to maximise reliability of measurement. To increase validity, assessors were interchanged between intervention and control groups after each clinic pair, to ensure that there was no systematic bias between the groups. The Mullen assessments were video recorded. As quality control, a random 10% sample of videos was reviewed internally and another random 10% sample was reviewed by an independent external assessor prior to the follow up survey.

The proportion of HIV exposed or infected children with full retention in care (> 80%) of scheduled visits at 12 months.

Secondary outcomes (Table 2)include HIV infected infants viral load measurement (> 1000 copies per ml at 12 months), ART adherence and retention in both HIV positive mothers and HIV infected infants, parental stress levels and mental health status of caregivers, household food security status and infant nutritional status.

Based on data collected in our pilot study, the mean Mullen ELC Score was 110.6 (standard deviation [SD] 16.1). The coefficient of variation (k) between clusters was 0.07 but this does not take account of the clustering effect of the group-based intervention so a value for k of 0.15 was used. Using these values, assuming a harmonic mean of 16 dyads enrolled per cluster and a loss to follow-up of 20%, 15 clusters per arm were needed to provide 80% power to detect an effect size (difference in means/SD) of 1.23 for the Mullen ELC Score. The same sample size provides 80% power to detect a risk difference of 20% in retention in care, assuming retention is 65% in the control arm and k = 0.2. Based on these calculations our recruitment target was 528 caregiver-child dyads in total from 30 clinics. We recruited 574 dyads to ensure a harmonic mean of 24 dyads was enrolled from the larger seven clinics and a harmonic mean of 12 dyads was enrolled from the smaller eight clinics. This was done to ensure ECS group sizes were sufficient to include 2 ISALS groups per ECS group.

The imbalance between arms was minimised by using restricted randomization, minimising on the number of HIV exposed infants between 0 and 24 months per clinic. Mapping of all clinic clusters within the two districts was conducted before recruitment and enrolment in each district. Of all the mapped clinics, those with fewer than 30 annual deliveries were excluded from the randomization process. The remaining 30 clusters were randomly allocated to the intervention versus usual care arms. The intervention was delivered to groups of 12 dyads, and 6 health facilities of sufficient size to run one group and 9 facilities of sufficient size to run two groups were allocated to each arm (Fig. 2). To maximise transparency and buy-in from key stakeholders, a public randomisation procedure was undertaken (in January 2016 in the first district and in May 2016 in the second district) involving MOHCC, and district level governance and medical representatives.

Blinding of patients and programme implementers were not possible at baseline because the participants had to know whether they were receiving the ECS programme. At 12 months follow up, the assessments was conducted outside trial communities with participants, from control and intervention communities within the same district, seen together. The assessor was blind to community allocation. Otherwise, data collection procedures were identical.

During the enrolment visit, all participating caregiver-child dyads were allocated a unique identifier to ensure patient confidentiality. Baseline data were collected using an interviewer-administered questionnaire and Audio Computer Assisted Survey Instrument (ACASI) for the sensitive parts of the questionnaire administered with the caregiver and a developmental assessment of the child. Caregivers received guidance on using ACASI from trained surveyors with regards to using the laptop and headphones to listen to instructions, questions, and responses that have been digitally recorded onto the ACASI platform.

The questionnaires collected information on the following domains: demographic characteristics, household characteristics, income and expenditure, food security, antenatal, delivery and post-natal care related to birth of participating child, infant and child feeding practices, maternal mental health, parental stress, HIV testing, disclosure and treatment history for self and participating child, self-reported ART adherence. Questionnaire data were entered directly onto tablets pre-programmed using Open Data Kit with range and consistency checks incorporated.

A follow up assessment was conducted among enrolled caregiver-child dyads after 12 months of programme implementation. This assessment was conducted in the same order in which the clusters were enrolled. Caregivers completed the follow up questionnaire administered in the same way as at baseline. Children had a full developmental assessment conducted by a trained assessor.

CeSHHAR Zimbabwe was the data-coordinating centre. In the field, data were uploaded to cloud storage daily and on password-protected office servers at the end of each week. The server was only accessible to the project data manager and named study personnel, on a central computer. Other hard-copy data were stored separately and securely in the field and then locked in a secure room at the data management centre. All data were cleaned, entered, analysed and stored.

Given the small number of clusters in the trial, cluster-level summary methods will be used. Analysis will be intention-to-treat. Age standardised ELC scores will be calculated from a combination of scores in the four cognitive domains of the Mullen Scales at endline using a HIV negative population in this setting, or a USA reference population if a local reference population cannot be obtained. The mean age-standardised composite score for each cluster will be estimated and the difference in means between arms calculated. The 95% CI for the mean difference will be estimated using a stratified t-test, with variance estimated from the residual mean square from an analysis of variance of cluster-specific means on stratum and trial arm. Analysis will adjust as fixed effects for baseline ELC score, stratum, and the following factors if they are imbalanced between arms at enrolment: age of children and caregivers, HIV status of children, socio-economic status and caregivers’ mental health. For the HIV-specific primary outcomes (proportion of children retained in care at 12 months and proportion with unsuppressed viral load) and for binary secondary outcomes, the unadjusted prevalence ratio will be estimated as the ratio of the geometric mean percentage in the intervention cluster versus the control cluster and compared using logistic regression. The main analysis will be complete case, with multiple imputation as a sensitivity analysis for the two primary outcomes, using variables associated with the outcome and with missingness. Further sensitivity analysis will use individual-level logistic regression and compare the result with the cluster-level methods described above.

The trial steering committee (principal investigators and investigators) will consider data sharing applications for quantitative and qualitative data. All applications for data sharing must be made through the principal investigator using a standard Data Sharing Form. The trial steering committee convenes a meeting to consider the application. There will be a good reason for turning down a request. Requests will be considered within 4 weeks of being made, and a decision communicated to the applicant as soon as possible thereafter.

The process evaluation integrates quantitative and qualitative data collection tools. In addition, key contextual factors that might affect whether the intervention will work in other settings in Zimbabwe were documented. Project documentation will be reviewed to assess whether activities were conducted as scheduled, to identify any delays or gaps, and to check for standardised comparison across intervention sites. Routine programme data were used to track participation in the intervention, complemented by qualitative research, using semi-structured in-depth interviews (IDIs) with ECS staff, case managers, clinic staff and participating caregivers conducted at 3, 6 and 12 months after the start of the programme. There were 24 CCWs, 15 community based trainers (trainers for the ISALS), 17 ECS facilitators, and 13 nurses involved in implementing the intervention.

The research team also conducted an in-depth process evaluation of the ISALS/ECS and case management programme. The process evaluation included gathering general information on programme implementation processes in intervention communities. Participant groups were randomly observed during ISALS collection / ECS sessions to provide data on programme fidelity. Project staff carry an “events diary” to make note of events that were beyond the control of the project, yet have the potential to influence implementation or outcomes (natural disasters, introduction of other ISALS/ECS services, and change in MOHCC standard of care).

IDIs with caregivers were conducted in the intervention arms at two intervals across the project (mid-implementation and end). Caregivers for the IDIs were purposively selected for specific attributes such as high or low levels of engagement in activities, users and non-users of clinical service. These interviews explored perceptions of the community programme, perceptions of clinical and other available services, and positive and negative experiences of the intervention. Semi-structured interviews, with health care staff at intervention clinics and the implementing stakeholders, were conducted to elicit perceptions of the acceptability of the intervention and their own levels of satisfaction and perspectives on its quality.

Ethical approval has been granted by the Medical Research Council of Zimbabwe (MRCZ/A/1943), Research Council of Zimbabwe, University College London (6789/002) and London School of Hygiene and Tropical Medicine (9912). The exposed Infant register was utilised within the study. This is not a research tool. The register is part of programme data under the Zimbabwe MOHCC data provision. The data are not publicly available and any requests to use that data are made directly to the Zimbabwe MOHCC services and considered on a case-by-case basis.

Any deviation from, or changes of, the protocol will be communicated to relevant partners and funding bodies via email. Changes to the current protocol at the Pan-African Clinical Trials Registry will be updated online.

At the time of the manuscript submission, recruitment for the trial was completed and process evaluation and data analysis ongoing.